EVALUATION OF LANDSLIDE SUSCEPTIBILITY IN THE TROPICAL MOUNTAINOUS REGION OF SINDANGKERTA, KABUPATEN BANDUNG BARAT, WEST JAVA

• Main Author: Dwi Kartiko, Rendy
• Format: Theses
• Language: Indonesia
• Subjects: Geologi, hidrologi & meteorologi
• Online Access: https://digilib.itb.ac.id/gdl/view/73645
• Institution: Institut Teknologi Bandung

Landslide occurrences in Indonesia are various phenomena. Landslides have been exhaustive problems mainly in roads, building structures, and agricultural fields. Landslide controlling factors in Indonesia area also has wide range of variety due to its complex geological setting, in combination with its various morphological conditions. This thesis analyzes landslide susceptibility in tropical mountainous region of Sindangkerta, Kabupaten Bandung Barat, West Java. The studied area located ±20 km southwest of Bandung City, and administratively belong to the Sindangkerta district. The studied area is dominantly part of Bandung Physiographic zone in the middle part of Western Java Island. Young age of Pliocene rocks dominantly cover the studied region. Beser Formation spread in southern part of the studied area, while Tertiary rocks stretched near Saguling Dam in the northern area. This study used several data layers for landslide susceptibility analysis such as morphometric data, soil and lithology data, buffer zone data, and landslide distribution. Bivariate statistics, logistic regression, discriminant analysis, SHALSTAB, and SINMAP methodology are used to generate susceptibility maps. Pixel size of 25x25 m is used for unit of analysis. Evaluation of susceptibility maps are in turn carried out with correlation matrix, Seed Cell Area Index (SCAI), and cumulative distribution with Area Under Curve (AUC). Bivariate statistics methodology has good performance for predicting landslide susceptibility in studied area. Multivariate statistics methods are good in predicting which parameter has more control over landslide occurrence. The main parameters controlling landslide occurrence are slope gradient, lithology, land use, and clay ratio. Statistical analyses in this study have higher performance than deterministic analyses as seen in AUC values. Data uncertainty and unit scale (25x25 m per pixel) in deterministic analyses are main cause for the low performance.